Apparently, though unproven, at 22:18 on Wednesday 10 November 2010, Dale did opine thusly:
> Dale wrote: > > According to top, gkrellm and cat /proc/meminfo there is no swap in > > use. I have 2Gbs of ram and have swappiness set to 20 or 30. I > > rarely use swap unless I am compiling something huge, OOo comes to > > mind, or have a LOT of images open with GIMP. > > > > I did check to make sure tho. My swappiness did get magically changed > > once before. I wish it was something that easy tho. > > > > Still open to ideas. I started a emerge -e world. > > > > Dale > > > > :-) :-) > > Just to update here. I started a emerge -e world. It has not even > finished yet but it appears to be working fine now. It was working > yesterday, last night, this morning and was working fine when I tried > just a minute ago. So, it appears that something needed to be > recompiled somewhere but no clue what that could have been. > > I'll keep testing over the next few days and may report back if it is > still working correctly. I hope that it does tho. It was getting on my > nerves. Useful tip to keep in mind: Sometimes emerge -e world works out great. It's way overkill mostly but unlike a sledgehammer to kill a mosquito, doesn't break the wall as well as kill the insect :-) IIRC, revdep-rebuild came about from the same line of thought. Some libs were being wrongly linked or linked to missing stuff and it was a huge ball-ache to find them all. Imagine running ldd on every binary and grepping for "not found" :-) It might even have been a glibc update (memory weak this end). revdep-rebuild finds the easily detectable stuff. But there's other problems that can happen with binaries that are not so easy to check (or not known to the dev), and none of the Gentoo tools help locate the culprit. emerge -e world will just rebuild everything in sight with the nice side effect of taking care of these mysterious problems. Hello sledgehammer. Pity that it can't record what it fixed though. It's interesting to see why Ubuntu and other binary distros never have this problem. First, they don't rip foundation libs out underneath a running system and insert different ones on the fly, and the API/ABI of their libs doesn't change for the life of that release of the distro. Plus, their build farms that generate new rpms/debs/pkgs nightly, essentially do the equivalent of a full emerge -e world daily and copy the binaries to the download server So sometimes when all else fails and suicide seems attractive, this is a workable approach that can help. Now if we can just get the gcc upgrade docs changed to reflect intelligent reality, we can get newbies to grok that emerge -e world is not suitable for the *first* fault-finding tool one uses.... -- alan dot mckinnon at gmail dot com
